On 04/08/14 09:34, Arthur Vaïsse-Lesteven wrote:
Hi everybody,
I'm still trying to use Jena Spatial and got a strange behavior.
I'm running a set of query to count all the entity of a given class in a radius
of a certain point.
For example, there are 3 interesting entity - ent1,ent2,ent3- and 3 classes -
cl1,cl2,cl3-.
I search for every entity of type cl1 near to ent1, then near to ent2, and last
near to ent3.
Then I do it for cl2 etc...
My problem is that a certain point (always the same), when spatial matching
become possible, it seems that JENA spatial is doing something wrong, and
browse all the index many times, looking for something. A debug logging extract
will follow ([1]).
Any idea of what can cause such a loop ?
The query engine does not have much information about the spatial index
so it does not attempt to reorder a spatial index access. It will be
executed as given.
ARQ uses index joins for BGPs so the first part
> ?entity a %class% .
> ?entity a <http://linkedgeodata.org/meta/Node> .
evaluates to a number of ?entity then it's a loop:
for each possible value E of ?entity
replace ?entity by E
E spatial:withinCircle (%entity_latitude% %entity_longitude%
%distance% 'm') .
evaluate
It might be better to put the spatial access first:
?entity spatial:withinCircle
(%entity_latitude% %entity_longitude% %distance% 'm') .
?entity a %class% .
?entity a <http://linkedgeodata.org/meta/Node> .
Andy
The queries that I execute are :
#Query that return the entity of class %class% in a radius of %distance%
#of the coordinate %entity_latitude% %entity_longitude%
PREFIX spatial: <http://jena.apache.org/spatial#>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
SELECT ?entity
WHERE{
#simple entity here. i.e. entity that are point.
{
?entity a %class% .
?entity a <http://linkedgeodata.org/meta/Node> .
?entity spatial:withinCircle (%entity_latitude% %entity_longitude%
%distance% 'm') .
}
UNION
#complex entity here
{
?entity a %class% .
?entity <http://geovocab.org/geometry#geometry> ?geometry .
?geometry <http://linkedgeodata.org/ontology/posSeq> ?sequence .
?sequence ?blankPredicate ?nodeGeometry.
?node <http://geovocab.org/geometry#geometry> ?nodeGeometry .
?node a <http://linkedgeodata.org/meta/Node> .
?node spatial:withinCircle (%entity_latitude% %entity_longitude%
%distance% 'm') .
}
}
where %class% is the class of searched entity (ex: cl1), %entity_latitude%
%entity_longitude% are the lat/long properties of the entity of interest (ex:
ent1 lat and long) and, obviously, %distance% is the radius of the search.
I execute such queries on data coming from Linked Open Geo
Data(http://linkedgeodata.org)that are conversion of Open Street Map data to
RDF.
Here comes an excerpt of the log file :
2014-08-04 09:23:49,407 [DEBUG] (ReorderTransformationSubstitution.java:reorder:88) Reorder: (?entity
<::type> <::AbandonedRailway>) (?entity <::type> <::Node>)
2014-08-04 09:23:49,407 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:144) >> Input
0 5 : ?entity <::type> <::AbandonedRailway>
1 5 : ?entity <::type> <::Node>
2014-08-04 09:23:49,407 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:153) << Output
?entity <::type> <::AbandonedRailway>
2014-08-04 09:23:49,407 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:144) >> Input
0 : null
1 1 : TERM <::type> <::Node>
2014-08-04 09:23:49,407 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:153) << Output
TERM <::type> <::Node>
2014-08-04 09:23:49,407 [DEBUG] (ReorderTransformationSubstitution.java:reorder:88) Reorder: (?entity
<::type> <::AbandonedRailway>) (?entity <::geometry> ?geometry) (?geometry
<::posSeq> ?sequence) (?sequence
?blankPredicate ?nodeGeometry) (?node <::geometry> ?nodeGeometry) (?node <::type>
<::Node>)
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:144) >> Input
0 5 : ?entity <::type> <::AbandonedRailway>
1 30 : ?entity <::geometry> ?geometry
2 30 : ?geometry <::posSeq> ?sequence
3 100 : ?sequence ?blankPredicate ?nodeGeometry
4 30 : ?node <::geometry> ?nodeGeometry
5 5 : ?node <::type>
<::Node>
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:153) << Output
?entity <::type> <::AbandonedRailway>
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:144) >> Input
0 : null
1 2 : TERM <::geometry> ?geometry
2 30 : ?geometry <::posSeq> ?sequence
3 100 : ?sequence ?blankPredicate ?nodeGeometry
4 30 : ?node <::geometry> ?nodeGeometry
5 5 : ?node <::type> <::Node>
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:153) << Output
TERM <::geometry> ?geometry
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:144) >> Input
0 : null
1 : null
2 2 : TERM <::posSeq> ?sequence
3 100 : ?sequence ?blankPredicate ?nodeGeometry
4 30 : ?node
<::geometry> ?nodeGeometry
5 5 : ?node <::type> <::Node>
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:153) << Output
TERM <::posSeq> ?sequence
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:144) >> Input
0 : null
1 : null
2 : null
3 10 : TERM ?blankPredicate ?nodeGeometry
4 30 : ?node <::geometry> ?nodeGeometry
5 5 : ?node <::type> <::Node>
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:153) << Output
?node <::type> <::Node>
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:144) >> Input
0 : null
1 : null
2 : null
3 10 : TERM ?blankPredicate ?nodeGeometry
4 2 : TERM <::geometry> ?nodeGeometry
5 : null
2014-08-04 09:23:49,408 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:153) << Output
TERM <::geometry> ?nodeGeometry
2014-08-04 09:23:49,409 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:144) >> Input
0 : null
1 : null
2 : null
3 10 : TERM ?blankPredicate
TERM
4 : null
5 : null
2014-08-04 09:23:49,409 [DEBUG]
(ReorderTransformationSubstitution.java:chooseNext:153) << Output
TERM ?blankPredicate TERM
2014-08-04 09:23:49,779 [DEBUG]
(SpatialOperationWithCircleBase.java:objectToStruct:178) Trying SpatialMatch:
(Circle(Pt(x=-0.078548,y=51.52302), d=0.0° 0.10km) -1 IsWithin)
2014-08-04 09:23:57,118 [DEBUG]
(SpatialOperationWithCircleBase.java:objectToStruct:178) Trying SpatialMatch:
(Circle(Pt(x=-0.078548,y=51.52302), d=0.0° 0.10km) -1 IsWithin)
[...]
2014-08-04 09:27:24,324 [DEBUG]
(SpatialOperationWithCircleBase.java:objectToStruct:178) Trying SpatialMatch:
(Circle(Pt(x=-0.078548,y=51.52302), d=0.0° 0.10km) -1 IsWithin)
The SpatialOperationWithCircleBase is executed 38 times before going to the
next step. It takes near to 20 minutes to get a result for a simple spatial
query with the current implementation, I assume that this isn't normal.
Thanks in advance,
VAISSE-LESTEVEN Arthur.
